- 3P Sampling
- 3P = probability proportional to prediction
- probability of a tree chosen for sampling (i.e. measured)
- is proportional
- to its predicted size
- restated
- the bigger it is ...
- the more likely it will be sampled
- Basics
- even people with limited experience
- can estimate size fairly consistently
- accuracy is NOT important
- consistency is the KEY!!
- advantage is high precision
- really it's low variability (CV)
- typically CV for cruising ~60%
- for 3P ~20%
- means fewer plots to get a "good SE"
- overview of field work
- go to every individual and guess its size
- compare est. size to a random number (more on this later)
- if EGER then measure CAREFULLY
- overview of compilation
- sum all est. values
- but we don't know "how good" our estimates are
- so we need a correction ratio (R)
- R = measured value / est. value
- R then "corrects" our estimate
- Real Total = Est. Total * correction Ratio
- Key Points
- need to visit each individual (tree)
- so estimate needs to be QUICK
- consistency important NOT accuracy
- due to high precision ... only a FEW samples needed ...
- ... need to be measured carefully
- for timber cruising
- traditional 3P restricted to
- small areas or corridors
- marking cruise (selection cuts)
- 100% cruise is required
- modified 3P sampling for larger areas
- Example ... Allan and his BYL
- Planning
- Some Terms
- ∑KPI ... is the total of the estimates
- (K+Z) ... is the maximum random number
- Calculate Sample Size
- n = CV^2 * t^2 / E%^2
- in order to calculate we need to know
- estimated CV
- confidence level - it determines which col. in t-table
- acceptable E% (i.e. 15%?)
- Create the Random Number Table
- remember EGER
- thus the random number table determines when to sample
- let's talk about chance
- random numbers 1-100
- chance of being selected?
- est. tree size is 20
- est. tree size is 50
- likely sample size?
- 8 trees, each is est. to be 50
- 8 trees (20, 40, 75, 10, 30, 15, 60, 55)
- equation for likely sample size is ...
- n = ∑KPI / (K+Z)
- emphasize it is the LIKELY sample size (n)
- Calculate (K+Z)
- rearrange n = ∑KPI / (K+Z) ...
- ... (K+Z) = ∑KPI / n
- Once again, but in order
- Planning
- determine sample size
- est. CV
- confidence level (95%?)
- acceptable error (15%?)
- determine (K+Z)
- est ∑KPI
- desired n (from above)
- generate the random # table
- calculator (or Excel) to get RAND# (0 - 1)
- multiply RAND# by (K+Z)
- Field
- go to each individual (tree) and est. size
- if EGER then carefully measure
- Compilation
- Actual Total = Total of estimates * correction ratio = ∑KPI * R
- Calculate
- ∑KPI
- individual R's and ave. R
- Vol = ∑KPI * ave R
- stats (CV%, SE% & E%) based on individual R's
- E% * Actual Total = total E in units ... provides confidence interval